Products

Agents

Claude agents are role definitions that let Claude run a whole job end to end: a brief goes in, finished work comes out. The 18 agents here were built and tested inside a working agency.

An agent here is not a prompt with a new name. Each one has a defined role, clear limits on what it will and will not do, and a hand-over rule: when it finishes its part, it tells you which agent should take the work next. That is how several agents run one job together, the same way people on a team pass work between desks.

In Claude Code these are the agents (or subagents) you define as role files and dispatch from the main thread; every agent on this shelf ships in exactly that format. The role definitions are plain text and open, so any AI can run them: native in Claude, adapted on any MCP-compatible stack.

18 in the catalog
Who this shelf is for
  • Founder who wants whole jobs taken off their plate, not more tools to assemble
  • Agency where work keeps falling apart as it passes from step to step
  • Developer who tried single prompts that work in a demo but stall on real work
Where to start

Start with one agent in the area that hurts most: content, growth or engineering, and run it on its own for a week. Once you trust its output, add the agents it works with; the agent kits package the groups that are proven to work together.

New to this shelf? Read the guides first: Claude Code Agents and Subagents Explained

Filter by category
Data & Analytics Agent

Data Weaver

Supabase/Postgres schema, RLS and API engineer

$79
Inspect →

Works with

Universal by design: these run in any AI. Delivered in the open Agent Skills + MCP format (native in Claude); ChatGPT, Gemini, Cursor and Copilot adapt the same files their own way.

  • Claude Native format
  • ChatGPT Adapts via open standards
  • Gemini Adapts via open standards
  • Cursor Adapts via open standards
  • Copilot Adapts via open standards
Questions · still in the air

Catch what's on your mind.

the air is clear. nothing between you and the forge.
catch a spark: the forge will answer

  1. What makes an agent different from a skill?

    A skill is a capability; an agent is a specialist. The agent carries its role, its rules, its memory and its hand-off discipline: you give it the job, it runs the whole lane end to end and returns finished work.

  2. Do agents run on my machine or yours?

    On yours: inside your own Claude environment, with your keys and your data staying with you. The one exception is Vorkaz, the licensed operator that runs on our servers and hands back results.

  3. Can the agents work together on one job?

    That is how they were built. Each agent ends by naming who should run next, the same chain discipline our agency uses: the builder hands to the SEO reviewer, the reviewer to QA. You can run one alone or wire the relay.

  4. Do agents only run in Claude?

    No. An agent is a role file in plain text, so it is not locked to one assistant. Claude Code runs it natively from the main thread; the same definition adapts to any other MCP-compatible AI stack.

  5. Can I run an agent with ChatGPT, Gemini or Cursor?

    Yes. The role, rules and hand-off are written as open plain text, so another MCP-compatible AI can run the same agent its own way. The one exception is Vorkaz, the licensed operator that runs on our servers and hands back results.